wordcount

name	wordcount
description	Count words in the LaTeX project using texcount — total across all included files plus a per-section breakdown — reading the main file from .latex-agentic.json. Falls back to a conservative detex-style pipeline when texcount is unavailable, and explains what is and isn't counted. Use when the user says "word count", "how many words", "wordcount", "count words", "how long is this".

wordcount

Report the word count of the document, following \input/\include files, and break it down by section.

Procedure

1. Read the project contract for the main file. From .latex-agentic.json, read main (fall back to main.tex):

   MAIN=$(jq -r '.main // "main.tex"' .latex-agentic.json 2>/dev/null || echo main.tex)
   

2. Total count (preferred — texcount). -inc follows included files, -total prints just the grand total, -brief keeps output tidy:

   texcount -inc -total -brief "$MAIN"
   

   Report the total word count. texcount separates text words from header/caption words and counts of math — surface the headline "words in text" number plus, when useful, words in headers and captions.

3. Per-section table. Re-run texcount split by section to show where the words are:

   texcount -inc -sub=section "$MAIN"
   

   Summarize as a small table: section title → words in text (and headers/captions/math if the user wants detail). This makes it obvious which section is over- or under-weight.

4. Fallback when texcount is missing. If command -v texcount fails, use a conservative strip-and-count pipeline. Strip comments, drop the preamble, remove commands and math, then count words:

   # Concatenate main + any \input/\include is best-effort here; run per file.
   sed -e 's/\([^\\]\)%.*/\1/' -e 's/^%.*//' "$MAIN" \
     | awk '/\\begin\{document\}/{p=1;next} /\\end\{document\}/{p=0} p' \
     | sed -e 's/\$[^$]*\$//g' \
           -e 's/\\[a-zA-Z@]*\*\?//g' \
           -e 's/[{}\[\]]//g' \
     | tr -s ' \t' '\n' \
     | grep -E '[[:alpha:]]' \
     | wc -w
   

   This is an approximation: it removes inline math and command tokens but cannot follow \input automatically, so run it per source file and sum, and clearly label the result as approximate. Prefer installing texcount (tlmgr install texcount) for an accurate figure.

5. Explain what counts. Always tell the user what the number includes so they can interpret it:

   • Body text words are the headline figure.
   • Headers (section/chapter titles), captions, and the abstract are counted by texcount but reported in separate buckets — they can be toggled. Note that journal/word limits usually mean body text only.
   • Math is reported as inline/displayed counts, not words.
   • Footnotes are included in the text count; the bibliography is not counted.
   • The fallback pipeline is a rougher estimate and excludes included files unless run on each.